LLM Driven Slide Generation
LLM-driven slide generation is an automated approach to creating presentation decks by leveraging large language models to convert structured text—typically markdown—into formatted slides. Rather than manually authoring slides through traditional presentation software, this approach uses LLMs to interpret content and generate appropriate slide layouts, transitions, and formatting. Tools like Marp, a markdown-based presentation ecosystem, serve as the rendering engine that converts the generated markdown into visual presentations.
Workflow and Process
The typical workflow involves providing source content to an LLM, which then generates markdown following Marp’s syntax conventions. The LLM handles structuring the content into logical slide divisions, determining appropriate heading hierarchies, and formatting code blocks, lists, and other elements suitable for presentation. The generated markdown is then processed by Marp to produce the final presentation, typically as HTML or PDF output.
Technical Architecture
The system combines three primary components: the LLM that performs content interpretation and markdown generation, the markdown specification (usually Marp dialect) that defines slide structure, and the rendering engine that converts markdown into visual output. This separation of concerns allows flexibility in swapping LLM providers while maintaining consistent markdown output, and enables version control of presentations as text files rather than proprietary binary formats.
Applications and Limitations
This approach is effective for generating presentations from structured input like technical documentation, lecture notes, or standardized content templates. The main limitations include challenges with complex visual layouts that markdown cannot easily express, dependency on LLM quality for content organization, and the requirement that source material be sufficiently structured for meaningful automation.